Overestimating global ecosystem respiration by Earth System Models under future warming

The land carbon cycle-climate change (C4) feedback, which partially determines the future level of climate warming, depends on the temperature sensitivity of ecosystem respiration (ER). Most of Earth System Models (ESMs) predict a strong C4 feedback by using monotonic response functions, whereas recent empirical evidence strongly suggests that ER does not monotonically increase with temperature. Here we used a Data-Informed Ecosystem Respiration Model (DIERM) to estimate global ER, and found that ESMs in Coupled Model Intercomparison Project Phase 6 (CMIP6) CMIP 6 (i.e., Can-ESM5, CESM2-WACCM, CMCC-ESM2, MPI-ESM1-2, and Nor-ESM2) generally overestimate ER in places with high air temperatures (e.g., tropical and temperate regions). Moreover, the overestimation of ER by ESMs increases with the increasing air temperature under future climate scenarios. Compared with our data-driven approach, Can-ESM5, CESM2-WACCM, CMCC-ESM2, MPI-ESM1-2, and Nor-ESM2 over-estimated global ER by 98.7%, 45.0%, 31.5%, 51.4%, and 64.8%, respectively, under the SSP585 scenarios by 2100. Overall, this study highlights the importance of accounting for the unimodal (functions with one maximum) temperature response pattern on ER and suggests that current models do not accurately represent the response of ER to warming, which may contribute to the large uncertainty of projected warming in the future.